Files in this item



application/pdfmachine-learnin ... erspective-aspublished.pdf (770kB)
.pdf of article as published in PMLAPDF


Title:Machine Learning and Human Perspective
Author(s):Underwood, Ted
Machine learning
Literary theory
Speculative fiction
Science fiction
Fantasy fiction
Digital humanities
Abstract:We've been taught that numbers are good at measuring objective facts, and ham-handed when it comes to slippery subjective questions. But this folk maxim is exactly wrong about twenty-first-century quantitative methods. Machine learning algorithms are actually quite bad at being objective, and very good at absorbing human perspectives implicit in the evidence used to train them. They will be valuable for humanists, not as objective oracles, but as ways of representing parallax and historical change. To dramatize perspectival uses of machine learning, I train models of genre on groups of books categorized by historical actors who range from Edwardian advertisers to contemporary librarians. Comparing the perspectives implicit in their choices casts new light on received histories of genre. Scientific romance and science fiction—whose shifting names have often suggested a fractured history—turn out to be more stable across two centuries than the genre we call fantasy.
Issue Date:2020-01-20
Publisher:Modern Language Association of America
Citation Info:Ted Underwood, "Machine Learning and Human Perspective," PMLA 135.1 (Jan 2020): 92-109.
Sponsor:Work on this article was supported by a Meyer H. Abrams fellowship at the National Humanities Center.
Rights Information:The publisher's exclusive license to distribute expired after one year; as of January 2021, this publication is in the public domain licensed CC-BY.
Date Available in IDEALS:2021-01-01

This item appears in the following Collection(s)

Item Statistics